The navitia developers have informed me that the "quality" of
auto-completion results is deprecated and already meaningless.
Also, we had a wrong inverse sorting order before. The only thing that
matters is the sort order in which the results are returned.
This commit fixes the sort order accordingly.
This commit adds the house number behind the street name if available
and gives a way to define other formats like in France
where the house number comes before the street name.
Navitia provides an "au" coverage which aggregates GTFS feeds from
most states public transport authorities. At time of writing, it supports:
* Melbourne
* Sydney
* Canberra
* Brisbane
* Darwin
* Alice Springs
* Perth
* Adelaide
* Hobart
* Launceston
* Bernie
However of these, Darwin, Alice Springs, and Darwin are out of date.
By default, only does a semi-comprehensive test of Melbourne (the PTV
network). This is to avoid running into Navitia's API limits if run
under CI, or run continuously during development.
For each other supported network (which is in date), it also provides
tests. The sum of all these tests is that each network should have each
different type of transport (e.g. train/tram/ferry/etc) tested.
This allows coverages with multiple distinct transport agencies
(e.g. Australia) to meaninfully distinguish between different lines
with similar names. For example, there is often buses which are
labelled by numbers. "Bus 12" needs to be two different colours,
based on which transport network it belongs to.
In the case of Australia, it will use this to delegate to the
lineStyles() method.
Use the `min_nb_journeys` parameter rather than the `count` parameter
when requesting trips.
The `count` parameter causes Navitia to remove trips arbitrarily
from the result which can look like a bug to the user
when she knows a trip should be there, but it isn't.